Automated Configuration of Component-based Distributed Real-time and Embedded Systems from Feature Models

Abstract

Component-based distributed real-time and embedded (DRE) systems facilitate the reuse of software artifacts across applications. To achieve a high-level of reuse, component-based DRE systems leverage late binding to allow dynamic system assembly at deployment time (e.g., via configuration scripts) rather than statically at compile-time. The complexity of deriving a correct manual configuration of an arbitrary set of components, however, is a key source of system failures, downtime, and missed deadlines. This paper presents a model-driven engineering tool called Fresh that uses (1) feature models to codify the configuration rules of components, (2) a constraint solver to derive a correct application configuration, and (3) an XML annotation engine to inject configuration decisions from the constraint solver directly into an application configuration. We use an avionics mission computing case study based on the Lightweight CORBA Component Model (CCM) to (1) demonstrate the complexities of configuring components in/out of a component-based application and (2) motivate the reduction in the configuration complexity when Fresh is used. The results show that Fresh achieves ~80-90% reduction in manual configuration effort, while also ensuring that the derived configuration is correct with respect to application configuration constraints and QoS requirements.